QTLNetwork_GPU
Software based on GPUs+CPUs computing for Mapping QTL with Epistatic and QE interaction effects in Experimental Populations
Futao Zhang , Xu Yang, Zhihong Zhu ,Jian Yang and Jun Zhu
Download
QTLNetwork_GPU is command-line software for mapping quantitative trait loci (QTL) with epistatic effects and QE interaction effects in DH, RI, BC, F2, IF2 and BxFy populations. A new GUI (Graphical User Interface) for QTLNetwork is under developing, and will be released soon. It is a hybrid computing software that exploits both GPUs computing and CPUs computing. The parallel computing parts of the software were developed by Futao Zhang in MSRA as an intern with the help of his mentor Xu Yang. This software is foucsed on the speed and achieved almost 30 times speedup than the CPU version with the data file which has 6000 observations and 165 markers.The QTLNetwork CPU version 2.2 was developed by Zhihong Zhu and 2.0 by Jian Yang.
QTLNetwork_GPU requires that your system be equipped with a NVIDIA CUDA-compatible device. The NVIDIA drivers must be version 257.21 (or greater) for Windows systems.
How to run
First make sure your system has CUDA GPUs and the CUDA driver is successfully installed. Then unzip the downloaded software and click the executable file QTLNetwork.exe.
If you are encountered with error 14001 - This application has failed to start because the application configuration is incorrect, you should install the VC library. For x64 systems, download vcredist_x64.exe. For x86 systems, download vcredist_x86.exe.
If you are encountered with Cuda error: the launch timed out and was terminated, you should disable the watchdog of Windows.
QTLNetwork_GPU 2.2β can run on any Windows systems with a NVIDIA CUDA-compatible device. But QTLNetwork_GPU2.0 only can run on the GPUs whose compute capability is 1.3 or higher.
Acknowledgement
This research was partially supported by
MSRA UR project.